Daryl Mahon, D.SoC, MA, Department of Applied Social Studies, National University of Ireland, Maynooth
George (Jeb) S. Brown, PhD, Center for Clinical Informatics
Ashley Simon, MA, Center for Clinical Informatics
Published on the Society for the Advancement of Psychotherapy
The purpose of this paper is to explore the relationship between alliance measures and improvement on the outcome measure for clients in longer term treatment (six to 20 sessions). This is the second of a two part series. The first article (Mahon et al., 2021) explored the relationship between alliance scores and improvement on the outcome measure for therapy lasting five sessions or less (56% of the sample available). While we extended the findings from previous meta-analysis (Flückiger et al., 2018) with a larger dataset, less is known about the alliance-outcome association in later sessions. For example, Baier et al. (2020) in their systematic review suggest that the alliance may be more important early on in treatment, as later sessions could be less likely to experience ruptures. If the alliance outcome association proves to follow different patterns at later stages of therapy than our previous findings, this will be important for practitioners to understand as it may highlight different junctures where the alliance needs more or less focus by clinicians .
Thus, this paper investigates the relationship between alliance scores, session by session change, and final outcome, for clients receiving over five sessions using data from the ACORN collaboration database (Brown et al., 2015a; Brown et al., 2015b). Specifically, we explore how the relationship between alliance scores and outcome may change depending on the length of treatment, six to 10 sessions, and 11 to 20 sessions.
Methods
QUESTIONNAIRES
This study utilizes a suite of questionnaires developed within the ACORN collaboration (Brown et al, 2015a). The first article in this series (Mahon et al., 2021) provides a more comprehensive description of the questionnaires employed.
All items use a 5-point Likert scale. The questionnaire subscales are scored as the mean of the non-missing items resulting in a range of 0-4. Higher scores indicate greater disturbance in the alliance.
SAMPLE
A total of 41,171 adult clients with intake scores in a clinical range who initiated treatment between 1/1/2017 and 12/31/2019. The average number of sessions was 6.7. Of these, 25,719 (70%) completed at least two assessments, with an average number of sessions of 9.4. Of this subset with multiple assessment, 14,134 (56%) completed treatment within five sessions. A total of 19,637 (76%) completed treatment within 10 sessions, and 23,050 (90%) completed treatment within 20 sessions.
RESULTS
The following analyses are based on the subsample completing treatment within five sessions. Results for longer lengths of treatment will be addressed in subsequent articles.
Alliance Descriptive Statistics
The alliance is notable in that it does not begin to approach a normal distribution, with scores heavily skewed towards a positive rating of alliance. At all sessions more than 75% rated the Alliance as highly as possible on the scale of 0-4, with 0 indicating no concerns re-alliance (perfect alliance). Likewise, mean session to session change is quite small.
Table 1
Alliance Change and Global Distress Scale Change
Table 2 presents the percentage of patients reported improved, same, or worsened from the prior session to the current session. As treatment progressed, the percentage of clients reporting no change increased while the percentage reporting either improved or worsened alliance decreased. However, when looking at change in alliance from the start of treatment (Table 3), the percentage showing change in either direction remains quite stable as the number of sessions increased.
Table 2
Table 4 presents Global Distress Scale (GDS) effect size from session to sessions as well as from the start of treatment as a function of direction of alliance change from the prior session.
Table 4
Table 5 presents the results of an analysis of variance (general linear model) evaluating the effect size based on the change in alliance since the start of treatment (better, same, worse). At each of the session ranges, a negative change in alliance was associated with a significantly lower effect size at that point in treatment. While the comparison failed to reach a significance level of p<.05, there was still a strong trend (p<.1), The failure to reach a higher level of significance is likely due to the shrinking sample sizes at latter sessions.
Table 5
Correlation Between Mean Alliance Scores and Change on Outcome Measure
The data set also permitted evaluation of mean alliance scores at difference points in treatment and improvement on GDS. Average alliance scores were calculated for four “waves” of sessions: two to five, six to 10, 11 to 15 and 16 to 20. Spearman’s Rho (Rs) correlation was used to evaluate the relationship between mean alliance scores in proceeding waves and GDS effect size at sessions 10, 15, and 20. For example, at session 10, mean alliance scores for sessions two to five and for six to 10 were correlated with effect size at session 10. Likewise, at session 20, the mean alliance for all for waves were correlated with the effect size at session 20. Table 6 presents the Spearman correlations (Rs) along with significance level and sample size at sessions 10, 15, and 20.
Table 6
At session 10, only the most recent wave of mean alliance scores was significantly correlated, and even then, the size of the correlations is quite small. At session 15, the second and third wave correlations were significant though again the magnitude of correlation was very modest (Rs=<.1). Curiously, at session 20, the first three waves of alliance scores were significant with Rs=>.1. The fourth wave of correlation (Rs=.09) failed to reach significance due in large measure to the small sample size.
In summary, alliance scores near then end of treatment tend to have stronger correlations to change on the outcome measures than to alliance scores earlier in treatment.
Alliance Presence or Absence and Global Distress Scale Change
The rate of missing alliance scores in these later sessions was quite low compared to those observed in short term treatment (up to five sessions). Unlike in short term treatment missing alliance scores at the later sessions were not significantly predictive of change at that session.
Summary
The results support the premise that alliance scores and change in alliance scores are predictive of improvement on the outcome measure at treatment episodes lasting more than five sessions. However, there is a diminishing effect as the number of sessions increases. This may in part be due to alliance scores overall improving as treatment progressed. Average alliance scores early in treatment were not predictive of change at later sessions. Only the average alliance scores for the most recent sessions were predictive of change up until that point, thus, possibly disconfirming the hypothesis offered by Baier et al. (2020), suggesting that the alliance may be critical early on in treatment, with alliance scores later in treatment less likely to fluctuate. However, we must note that the lack of alliance items being filled out did not seem to have the same implications on GDS scores than in our previous study of the first five sessions of therapy. It is possible that this reflects a strengthening therapeutic alliance in later sessions, which would be consistent with the hypothesis offered by Baier et al. (2020).
For clinicians, the implications of these results suggest that they should continue to collect alliance measures throughout the entire treatment episode. Despite good alliance and rapport early on in treatment, the clinician cannot assume it will remain the high throughout treatment. Consequently, it is important to track rapport throughout the relationship as it can help address any disruptions or changes which can lead to improved outcomes.